In recent years, competition authorities across the globe have ramped up their scrutiny of large internet service providers. Special focus has fallen on the potential risks posed by “big data”. Framing enforcement around big data means regulators prioritize the volume of consumer information a firm collects and controls, while attaching less weight to its market position within the internet services sector.

      However, John H. Chestnut Professor of Law, Communication, Computer and Information Sciences at the University of Pennsylvania, and the founding director of the University of Pennsylvania Center for Technology, Innovation and Competition, Christopher S. Yoo, however, pointed out that "big data" is increasingly resembling the enforcement slogans of law enforcement agencies against digital platforms, lacking the necessary empirical analysis and rational consideration for it. Professor Yoo intends to conduct an in-depth dissection of what “big” data truly means by illustrating the application of data in predictive analytics, reviewing empirical literature on data-driven economies of scale, and examining the significance of data attributes beyond mere volume.

      On September 5, 2019, at the seminar titled Institutional and Cultural Construction of the Digital Economy hosted by the Renmin Law and Technology Institute, Renmin University of China, Professor Christopher S. Yoo delivered a keynote speech entitled Beyond the Hype – Rethinking Antitrust and Digital Platforms: What Does “Big” Data Truly Mean?

      The following is a compiled record of Professor Yoo’s speech and the insightful guest commentaries for readers’ reference.

Keynote Speech Session:

      Today, my lecture will unfold the analysis along four dimensions. First, we need to expand the framework for understanding consumer data and focus on its actual economic value. Second, when evaluating data by Volume, we must also account for economies of scale and diminishing marginal returns inherent to big data. Third, different business models utilize data in distinct ways, so empirical analysis must be conducted on a case-by-case basis. Fourth, I will address market definition for targeted advertising and examine the competitive relationship between search engine markets and social media markets.

I. Multiple Dimensions of Data

      In the early days of big data, the term merely conjured up images of massive quantities of information. Over time, frameworks to describe data dimensions emerged, including the 3 Vs, 4 Vs (Volume, Velocity, Variety, Value), and even 6 Vs. Through continuous exploration and practice, our understanding of big data has grown deeper and more comprehensive. Volume refers to the sheer quantity of data; Velocity denotes the speed at which data is generated and updated; Variety means data comes from diverse sources; and Value stands for the actionable utility extracted from data.

      That said, I argue that data types are equally critical. The distinction between structured data and unstructured data exerts a substantial impact on data analytics.

      Therefore, any data analysis focusing solely on volume is one-sided, making it impossible to accurately assess the costs of data analytics and the impact of data on barriers to market entry. All data assessments should classify the relevant data types. Regulators ought to examine how firms deploy different categories of data before reaching proper evaluations, and exercise prudence when reviewing corporate acquisitions involving various forms of data.

II. Economies of Scale in Data

      What are data economies of scale? Take this lecture as an analogy: the more students attend, the greater the value of the lecture. Yet the classroom has a fixed capacity cap, so the lecture’s value cannot grow indefinitely — this illustrates diminishing marginal returns. The big data economy follows the same logic. Beyond recognizing data economies of scale, we must also identify the saturation point where additional data yields no further benefits. Simply put, more data is not always better.

      Drawing on a body of empirical analysis, we can at minimum conclude that the blanket claim of “big is bad” is an oversimplification. In other words, it lacks rigor to assume that firms such as FANG and GAFA possess identical market power merely because they hold large volumes of data. Accordingly, we need to devote greater research efforts to answering the subsequent question: what volume of data delivers peak operational efficiency?

III. Data Under Distinct Business Models

      Statistics from 2019 show that roughly half of all companies lack the analytical tools required to conduct productive, efficient data analysis. For data-driven enterprises, the absence of proper analytical tools and professional data analysts can prove nearly fatal. For one thing, massive stockpiles of raw data become completely useless. For another, flawed algorithms directly undermine corporate decision-making, day-to-day operations, and the quality of advertising recommendations.

      At present, there is no universal data extraction software available across the internet industry. As a result, data analysis remains heavily dependent on a company’s specific business model. E-commerce retailers, online advertisers and social networking platforms pursue distinct commercial objectives and adopt mutually incompatible proprietary algorithms.

      From a competition perspective, only companies with superior data processing technology can gain stronger competitive edges. Firms that merely hoard data without utilizing it or rely on defective analytical models do not merit antitrust scrutiny. Moreover, from an efficiency standpoint, certain practices that appear to constitute price discrimination can expand market access by attracting consumers with low willingness to pay, thereby boosting market competition and deserving regulatory encouragement.

IV. Market Definition for Targeted Advertising

      In my view, two considerations govern the definition of the relevant market for the advertising sector. First, regulators and courts should take advertisers’ perspectives into account when delineating the relevant market. If advertisers regard traditional advertising and digital advertising as interchangeable substitutes, the relevant market shall cover both categories. Conversely, if advertisers treat them as fundamentally distinct, the two ought to be separated into independent relevant markets. Second, the existence of substitutability must be verified by empirical evidence.

      To properly define the relevant market for advertising services, more empirical research is required to accurately assess and compare the marginal cost structures of traditional and digital advertising, which will clarify the competitive relationship between these two advertising markets. Nevertheless, existing empirical evidence suggests that ACCC’s conclusion that traditional and digital advertising are not substitutes for one another is likely flawed.

Conclusion

      We must conduct more rigorous and thorough research to avoid repeating past regulatory mistakes. All the empirical literature covering the four dimensions I discussed today centers on a single core inquiry: whether the collection and exploitation of data impairs market competition. Moving forward, empirical research should focus on the following four priorities:

(1) Multi-dimensional analysis of big data;

(2) The minimum data volume required to unlock effective value and the saturation threshold of data economies of scale;

(3) Human capital and technical investments essential to leveraging big data;

(4) Substitutability between online and offline advertising.

Comments and Q&A Session:

      During the comment and Q&A session, Professor Yang Dong, Vice Dean of the Renmin University of China Law School, raised questions concerning the identification of market power and the definition of market share. In response, Professor Christopher S. Yoo first briefly reviewed the legal rules governing market power in the United States, Europe and China. He noted that the U.S. legal standards for establishing a firm’s market power are comparatively stringent, leaving enterprises substantial autonomy over pricing practices. EU competition law contains the concept of "collective dominance", yet this legal provision rarely comes into play in actual law enforcement. Therefore, Professor Christopher S. Yoo emphasized that the assessment and definition of market power must proceed on a case-by-case basis, with separate examinations of each firm’s specific circumstances and its impacts on competitive markets. Furthermore, Professor Christopher S. Yoo also pointed out that market leaders have never remained static, whether in the past or at present. Even the currently dominant major internet giants are confronted with difficulties in their survival and long-term development.

      Other participating scholars carried out in-depth exchanges and discussions on relevant issues, including Professor Meng Yanbei, Associate Professors Zhang Jiyu, Guo Rui and Ding Xiaodong from the Renmin University of China Law School; Associate Professor Shen Weiwei from China University of Political Science and Law; and Researcher Fu Wei from JD Research Institute, among others.

Editor:Xu Liuya

Reviewer: Zhu Peiwen